Project details

Approach: Integration of data of different types and from multiple sources to design predictive analytical methods for optimization of flotation plant performance.

This project is integrating data from multiple sensors in a mineral processing plant with other nonsensor data obtained in the laboratory and other sources to build predictive models for optimization of a flotation operation, together with identifying key missing data needed to improve performance prediction and, importantly, rapid detection of adverse process trends and events. The approach is to examine, on a plant scale, with parallel, controlled laboratory measurements, model interrelationships between liberation variation, pulp chemistry and process outcomes (recovery and grade).

For plant studies, time series sampling, in parallel with pulp chemistry monitor (PCM) measurements is being undertaken, over several hours. Plant samples, for QEMSCAN liberation analysis and laboratory flotation tests (where appropriate). Fundamental laboratory studies are being undertaken, using controlled variations in value mineral using synthetic composites. The latter synthesis has been previously developed in work on composite particle flotation at the University of South Australia.

Initial analysis of liberation and pulp chemistry data (together with plant operating conditions, e.g. reagent addition, hydrodynamic variables) against plant/bank performance is being undertaken to ascertain liberation variability feeding flotation, sensitivity of plant performance (recovery, grade) and identifying the need for specific, additional sensor data. Integration of the fundamental liberation chemistry-performance test work is aiding in identifying limitations for detection (thresholds, significance) of plant data changes and trends. Various mathematical and statistical analysis tools are being employed to extract key interrelationships and trends in the data.

The project aims to design and develop effective and practical predictive optimization tools that can assist in rapid recognition and reaction to trend data. It is anticipated that outcomes will provide an important extension to research elsewhere in the Training Centre concentrated on specific sensing strategies for liberation.

Image courtesy of Magotteaux – the MagoPulp, measuring flotation chemistry.


Clement Lartey


University of South Australia

Key Contacts

Principal Supervisor:
Associate Professor Jixue Liu


Ph: +61 08 8302 3690

Professor William Skinner
Professor David Beattie
Professor Nigel Cook
Professor Marta Krasowska

Supervisor Profile


Lartey, C., Skinner, W., Asamoah, R., Greet, C., Zanin, M., Liu, J. 2023. Outlier Detection in Sensed Froth Flotation Data. Minerals Engineering.


Haavisto, O., Kaartinen, J. and Hyötyniemi, H. (2008). Optical spectrum-based measurement of flotation slurry contents. International Journal of Mineral Processing, 88(3-4): 80-88.

Industry Partners